Paperjet/backend/app/services/export/renderer.py
Elijah 4569dea864
Some checks failed
CI / Backend (Python) (push) Failing after 27s
CI / Frontend (TypeScript) (push) Successful in 14s
CI / Container (Docker) (push) Has been skipped
Implement Paperjet updates
2026-08-15 14:08:24 -07:00

979 lines
34 KiB
Python

"""Flatten canonical annotation data into a PDF with PyMuPDF."""
import logging
import math
import os
import re
import uuid
from collections.abc import Callable, Mapping, Sequence
from pathlib import Path
from typing import Any
import pymupdf
logger = logging.getLogger(__name__)
Annotation = Mapping[str, Any]
Handler = Callable[[pymupdf.Page, Annotation, Path], None]
def _rect(annotation: Annotation) -> pymupdf.Rect | None:
value = annotation.get("rect")
if not isinstance(value, Mapping):
return None
try:
x = float(value["x"])
y = float(value["y"])
width = float(value["width"])
height = float(value["height"])
except (KeyError, TypeError, ValueError):
return None
if width < 0 or height < 0:
return None
return pymupdf.Rect(x, y, x + width, y + height)
def _props(annotation: Annotation) -> Mapping[str, Any]:
value = annotation.get("props")
return value if isinstance(value, Mapping) else {}
def _color(
value: Any,
fallback: tuple[float, float, float] = (0.0, 0.0, 0.0),
) -> tuple[float, float, float]:
if not isinstance(value, str):
return fallback
text = value.strip().lstrip("#")
if len(text) == 3:
text = "".join(char * 2 for char in text)
if len(text) != 6:
return fallback
try:
return tuple(int(text[index : index + 2], 16) / 255 for index in (0, 2, 4)) # type: ignore[return-value]
except ValueError:
return fallback
def _opacity(value: Any, fallback: float = 1.0) -> float:
try:
return max(0.0, min(1.0, float(value)))
except (TypeError, ValueError):
return fallback
def _font_name(family: Any, bold: Any = False, italic: Any = False) -> str:
name = str(family or "Liberation Sans").lower()
is_serif = "times" in name or "serif" in name
is_mono = "courier" in name or "mono" in name
if is_serif:
normal, bold_name, italic_name, bold_italic_name = "tiro", "tibo", "tiit", "tibi"
elif is_mono:
normal, bold_name, italic_name, bold_italic_name = "cour", "cobo", "coit", "cobi"
else:
normal, bold_name, italic_name, bold_italic_name = "helv", "hebo", "heit", "hebi"
if bold and italic:
return bold_italic_name
if bold:
return bold_name
if italic:
return italic_name
return normal
_SIGNATURE_FONT_FILES = {
"great vibes": "GreatVibes-Regular.ttf",
"allura": "allura.ttf",
"sacramento": "sacramento.ttf",
"dancing script": "dancing-script.ttf",
"caveat": "caveat.ttf",
}
_WEB_FONT_FILES: dict[tuple[str, bool, bool], tuple[str, ...]] = {
("outfit", False, False): ("outfit-400.ttf",),
("outfit", True, False): ("outfit-700.ttf", "outfit-400.ttf"),
("plus jakarta sans", False, False): ("plus-jakarta-sans-400.ttf",),
("plus jakarta sans", True, False): ("plus-jakarta-sans-700.ttf", "plus-jakarta-sans-400.ttf"),
("plus jakarta sans", False, True): (
"plus-jakarta-sans-400-italic.ttf",
"plus-jakarta-sans-400.ttf",
),
("plus jakarta sans", True, True): (
"plus-jakarta-sans-700-italic.ttf",
"plus-jakarta-sans-700.ttf",
),
("great vibes", False, False): ("great-vibes-400.ttf", "GreatVibes-Regular.ttf"),
("allura", False, False): ("allura-400.ttf", "allura.ttf"),
("sacramento", False, False): ("sacramento-400.ttf", "sacramento.ttf"),
("dancing script", False, False): ("dancing-script-400.ttf", "dancing-script.ttf"),
("caveat", False, False): ("caveat-400.ttf", "caveat.ttf"),
}
_SYSTEM_FONT_FAMILIES = {
"arial": "LiberationSans",
"helvetica": "LiberationSans",
"liberation sans": "LiberationSans",
"courier": "LiberationMono",
"courier new": "LiberationMono",
"liberation mono": "LiberationMono",
"mono": "LiberationMono",
"serif": "LiberationSerif",
"times": "LiberationSerif",
"times new roman": "LiberationSerif",
"liberation serif": "LiberationSerif",
}
def _is_bold(value: Any) -> bool:
return value is True or str(value).strip().lower() in {"bold", "700", "800", "900"}
def _is_italic(value: Any) -> bool:
return value is True or str(value).strip().lower() in {"italic", "oblique"}
def _variant_filename(stem: str, bold: bool, italic: bool) -> str:
if bold and italic:
suffix = "-BoldItalic"
elif bold:
suffix = "-Bold"
elif italic:
suffix = "-Italic"
else:
suffix = "-Regular"
return f"{stem}{suffix}.ttf"
def _font_roots() -> tuple[Path, ...]:
source_root = Path(__file__).parents[2] / "assets" / "fonts"
configured_root = os.environ.get("PAPERJET_FONT_DIR")
roots = [Path(configured_root)] if configured_root else []
roots.extend(
[
source_root / "web",
source_root,
Path("/usr/share/fonts/truetype/liberation"),
Path("/usr/local/share/fonts"),
]
)
return tuple(dict.fromkeys(roots))
def _font_file(family: Any, bold: Any = False, italic: Any = False) -> Path | None:
normalized = str(family or "").strip().lower()
bold_flag = _is_bold(bold)
italic_flag = _is_italic(italic)
key = (normalized, bold_flag, italic_flag)
filenames = _WEB_FONT_FILES.get(key)
if filenames is None and italic_flag:
filenames = _WEB_FONT_FILES.get((key[0], key[1], False))
if filenames is None:
system_stem = _SYSTEM_FONT_FAMILIES.get(normalized)
filenames = (_variant_filename(system_stem, bold_flag, italic_flag),) if system_stem else ()
for root in _font_roots():
for filename in filenames:
path = root / filename
if path.is_file():
return path
return None
def _font_choice(family: Any, bold: Any = False, italic: Any = False) -> tuple[str, Path | None]:
font_file = _font_file(family, bold, italic)
if font_file is not None:
return f"paperjet-{font_file.stem.lower()}", font_file
return _font_name(family, _is_bold(bold), _is_italic(italic)), None
def _needs_synthetic_italic(family: Any, bold: Any = False, italic: Any = False) -> bool:
if not _is_italic(italic):
return False
normalized = str(family or "").strip().lower()
if not any(key[0] == normalized for key in _WEB_FONT_FILES):
return False
return (normalized, _is_bold(bold), True) not in _WEB_FONT_FILES
def _synthetic_italic_morph(origin: pymupdf.Point) -> tuple[pymupdf.Point, pymupdf.Matrix]:
# Match the browser's synthesized oblique for families without an italic face.
shear = math.tan(math.radians(14))
return origin, pymupdf.Matrix(1, 0, shear, 1, 0, 0)
def _signature_font_file(family: Any) -> Path | None:
normalized = str(family or "").strip().lower()
for filename in _WEB_FONT_FILES.get((normalized, False, False), ()):
for root in _font_roots():
path = root / filename
if path.is_file():
return path
legacy_filename = _SIGNATURE_FONT_FILES.get(normalized)
if not legacy_filename:
return None
path = Path(__file__).parents[2] / "assets" / "fonts" / legacy_filename
return path if path.is_file() else None
def _signature_font_size(
rect: pymupdf.Rect,
text: str,
font_file: Path | None,
font_name: str,
requested_size: Any = None,
) -> float:
if requested_size is not None:
try:
return max(1.0, float(requested_size))
except (TypeError, ValueError):
return 48.0
# Legacy typed signatures did not persist a font size. Preserve their
# existing sizing behavior while new records use the preview's canonical
# size above.
size = max(1.0, min(48.0, rect.height * 0.55))
try:
font = (
pymupdf.Font(fontfile=str(font_file)) if font_file else pymupdf.Font(fontname=font_name)
)
text_width = font.text_length(text, fontsize=size)
if text_width > rect.width and text_width > 0:
size *= max(0.1, rect.width / text_width) * 0.95
except Exception:
# A missing optional font must not prevent a PDF export.
pass
return max(1.0, size)
def _points_from_props(annotation: Annotation, rect: pymupdf.Rect) -> list[pymupdf.Point]:
props = _props(annotation)
raw_points = props.get("paths")
points: list[pymupdf.Point] = []
if isinstance(raw_points, Sequence) and not isinstance(raw_points, (str, bytes)):
for raw_point in raw_points:
if isinstance(raw_point, Sequence) and len(raw_point) == 2:
try:
points.append(
pymupdf.Point(
rect.x0 + float(raw_point[0]),
rect.y0 + float(raw_point[1]),
)
)
except (TypeError, ValueError):
continue
if len(points) >= 2:
return points
# Older clients stored a Fabric SVG path. Approximate its command end
# points and normalize them into the canonical annotation rectangle.
svg_path = props.get("svgPath")
if not isinstance(svg_path, str):
return []
raw = [float(value) for value in re.findall(r"[-+]?\d*\.?\d+(?:[eE][-+]?\d+)?", svg_path)]
if len(raw) < 4:
return []
pairs = list(zip(raw[::2], raw[1::2], strict=False))
min_x = min(point[0] for point in pairs)
max_x = max(point[0] for point in pairs)
min_y = min(point[1] for point in pairs)
max_y = max(point[1] for point in pairs)
source_width = max(max_x - min_x, 1e-6)
source_height = max(max_y - min_y, 1e-6)
return [
pymupdf.Point(
rect.x0 + (x - min_x) / source_width * rect.width,
rect.y0 + (y - min_y) / source_height * rect.height,
)
for x, y in pairs
]
def _draw_polyline(
page: pymupdf.Page,
points: list[pymupdf.Point],
*,
color: tuple[float, float, float],
width: float,
opacity: float,
) -> None:
if len(points) < 2:
return
shape = page.new_shape()
shape.draw_polyline(points)
shape.finish(
color=color,
width=max(0.1, width),
lineCap=1,
lineJoin=1,
closePath=False,
stroke_opacity=opacity,
)
shape.commit(overlay=True)
_SVG_PATH_TOKEN = re.compile(r"[A-Za-z]|[-+]?(?:\d*\.\d+|\d+\.?)(?:[eE][-+]?\d+)?")
_SVG_PATH_ARITY = {"M": 2, "L": 2, "Q": 4, "C": 6}
def _parse_svg_path(value: str) -> list[tuple[str, tuple[float, ...]]]:
tokens = _SVG_PATH_TOKEN.findall(value)
commands: list[tuple[str, tuple[float, ...]]] = []
command: str | None = None
relative = False
move_pending = False
index = 0
current = (0.0, 0.0)
subpath_start = current
while index < len(tokens):
token = tokens[index]
if token.isalpha():
command = token.upper()
relative = token.islower()
index += 1
if command == "Z":
commands.append(("Z", ()))
current = subpath_start
command = None
continue
if command not in _SVG_PATH_ARITY:
return []
move_pending = command == "M"
if command is None:
return []
arity = _SVG_PATH_ARITY[command]
if index + arity > len(tokens):
return []
if any(tokens[index + offset].isalpha() for offset in range(arity)):
return []
try:
values = tuple(float(tokens[index + offset]) for offset in range(arity))
except ValueError:
return []
index += arity
base_x, base_y = current
if command in {"M", "L"}:
point = (
values[0] + (base_x if relative else 0.0),
values[1] + (base_y if relative else 0.0),
)
output_command = "M" if move_pending else "L"
commands.append((output_command, point))
current = point
if output_command == "M":
subpath_start = point
move_pending = False
command = "L" if command == "M" else command
elif command == "Q":
control = (
values[0] + (base_x if relative else 0.0),
values[1] + (base_y if relative else 0.0),
)
endpoint = (
values[2] + (base_x if relative else 0.0),
values[3] + (base_y if relative else 0.0),
)
commands.append(("Q", (*control, *endpoint)))
current = endpoint
else:
control_one = (
values[0] + (base_x if relative else 0.0),
values[1] + (base_y if relative else 0.0),
)
control_two = (
values[2] + (base_x if relative else 0.0),
values[3] + (base_y if relative else 0.0),
)
endpoint = (
values[4] + (base_x if relative else 0.0),
values[5] + (base_y if relative else 0.0),
)
commands.append(("C", (*control_one, *control_two, *endpoint)))
current = endpoint
return commands
def _draw_svg_path(
page: pymupdf.Page,
rect: pymupdf.Rect,
svg_path: str,
*,
color: tuple[float, float, float],
width: float,
opacity: float,
) -> bool:
parsed = _parse_svg_path(svg_path)
if not parsed:
return False
segments: list[tuple[str, tuple[tuple[float, float], ...]]] = []
samples: list[tuple[float, float]] = []
current = (0.0, 0.0)
subpath_start = current
def add_line(start: tuple[float, float], end: tuple[float, float]) -> None:
segments.append(("L", (start, end)))
samples.extend((start, end))
def add_quadratic(
start: tuple[float, float],
control: tuple[float, float],
end: tuple[float, float],
) -> None:
segments.append(("Q", (start, control, end)))
for step in range(21):
t = step / 20
inverse = 1 - t
samples.append(
(
inverse * inverse * start[0]
+ 2 * inverse * t * control[0]
+ t * t * end[0],
inverse * inverse * start[1]
+ 2 * inverse * t * control[1]
+ t * t * end[1],
)
)
def add_cubic(
start: tuple[float, float],
control_one: tuple[float, float],
control_two: tuple[float, float],
end: tuple[float, float],
) -> None:
segments.append(("C", (start, control_one, control_two, end)))
for step in range(21):
t = step / 20
inverse = 1 - t
samples.append(
(
inverse**3 * start[0]
+ 3 * inverse**2 * t * control_one[0]
+ 3 * inverse * t**2 * control_two[0]
+ t**3 * end[0],
inverse**3 * start[1]
+ 3 * inverse**2 * t * control_one[1]
+ 3 * inverse * t**2 * control_two[1]
+ t**3 * end[1],
)
)
for command, values in parsed:
if command == "M":
current = (values[0], values[1])
subpath_start = current
samples.append(current)
elif command == "L":
endpoint = (values[0], values[1])
add_line(current, endpoint)
current = endpoint
elif command == "Q":
control = (values[0], values[1])
endpoint = (values[2], values[3])
add_quadratic(current, control, endpoint)
current = endpoint
elif command == "C":
control_one = (values[0], values[1])
control_two = (values[2], values[3])
endpoint = (values[4], values[5])
add_cubic(current, control_one, control_two, endpoint)
current = endpoint
elif command == "Z":
if current != subpath_start:
add_line(current, subpath_start)
current = subpath_start
if not segments or not samples:
return False
min_x = min(point[0] for point in samples)
max_x = max(point[0] for point in samples)
min_y = min(point[1] for point in samples)
max_y = max(point[1] for point in samples)
source_width = max(max_x - min_x, 1e-6)
source_height = max(max_y - min_y, 1e-6)
def map_point(point: tuple[float, float]) -> pymupdf.Point:
return pymupdf.Point(
rect.x0 + (point[0] - min_x) / source_width * rect.width,
rect.y0 + (point[1] - min_y) / source_height * rect.height,
)
shape = page.new_shape()
for command, points in segments:
mapped = tuple(map_point(point) for point in points)
if command == "L":
shape.draw_line(mapped[0], mapped[1])
elif command == "Q":
curve_start, curve_control, curve_end = mapped
mapped_control_one = pymupdf.Point(
curve_start.x + (curve_control.x - curve_start.x) * 2 / 3,
curve_start.y + (curve_control.y - curve_start.y) * 2 / 3,
)
mapped_control_two = pymupdf.Point(
curve_end.x + (curve_control.x - curve_end.x) * 2 / 3,
curve_end.y + (curve_control.y - curve_end.y) * 2 / 3,
)
shape.draw_bezier(
curve_start,
mapped_control_one,
mapped_control_two,
curve_end,
)
else:
shape.draw_bezier(*mapped)
shape.finish(
color=color,
width=max(0.1, width),
lineCap=1,
lineJoin=1,
closePath=False,
stroke_opacity=opacity,
)
shape.commit(overlay=True)
return True
def _style_at(styles: Any, line_number: int, character_number: int) -> Mapping[str, Any]:
if not isinstance(styles, Mapping):
return {}
line_styles = styles.get(str(line_number), styles.get(line_number))
if not isinstance(line_styles, Mapping):
return {}
character_style = line_styles.get(str(character_number), line_styles.get(character_number))
return character_style if isinstance(character_style, Mapping) else {}
def _insert_text_lines(
page: pymupdf.Page,
rect: pymupdf.Rect,
text: str,
*,
fontsize: float,
font_name: str,
font_file: Path | None = None,
color: tuple[float, float, float],
align: int = 0,
line_height: float = 1.2,
synthetic_italic: bool = False,
) -> None:
"""Insert text without requiring it to fit inside the saved rectangle.
Fabric keeps text visible even when a manually sized textbox is shorter
than the font metrics. PyMuPDF's textbox helper instead returns a negative
spare-height value and may insert nothing, so this fallback preserves the
preview's overflow behavior.
"""
font = pymupdf.Font(fontfile=str(font_file)) if font_file else pymupdf.Font(fontname=font_name)
for line_number, line in enumerate(text.split("\n")):
line_width = float(font.text_length(line, fontsize=fontsize))
x = rect.x0
if align == 1:
x += max(0.0, (rect.width - line_width) / 2)
elif align == 2:
x += max(0.0, rect.width - line_width)
options: dict[str, Any] = {
"fontsize": fontsize,
"fontname": font_name,
"color": color,
"overlay": True,
}
if font_file:
options["fontfile"] = str(font_file)
origin = pymupdf.Point(x, rect.y0 + fontsize + line_number * fontsize * line_height)
if synthetic_italic:
options["morph"] = _synthetic_italic_morph(origin)
page.insert_text(origin, line, **options)
def _render_styled_text(
page: pymupdf.Page,
rect: pymupdf.Rect,
text: str,
props: Mapping[str, Any],
styles: Any,
) -> bool:
if not isinstance(styles, Mapping):
styles = {}
base_size = max(1.0, float(props.get("fontSize", 14)))
line_height = max(0.5, float(props.get("lineHeight", 1.2)))
lines = text.split("\n")
y = rect.y0
for line_number, line in enumerate(lines):
measured: list[
tuple[
str,
float,
str,
Path | None,
tuple[float, float, float],
float,
str | None,
bool,
]
] = []
for character_number, character in enumerate(line):
style = _style_at(styles, line_number, character_number)
font_size = max(1.0, float(style.get("fontSize", base_size)))
font_family = style.get("fontFamily", props.get("fontFamily"))
font_name, font_file = _font_choice(
font_family,
_is_bold(style.get("fontWeight", "bold" if props.get("bold") else None)),
_is_italic(style.get("fontStyle", "italic" if props.get("italic") else None)),
)
synthetic_italic = _needs_synthetic_italic(
font_family,
style.get("fontWeight", "bold" if props.get("bold") else None),
style.get("fontStyle", "italic" if props.get("italic") else None),
)
color = _color(style.get("fill", props.get("color")))
font = (
pymupdf.Font(fontfile=str(font_file))
if font_file
else pymupdf.Font(fontname=font_name)
)
width = float(font.text_length(character, fontsize=font_size))
background = style.get("textBackgroundColor", props.get("highlightColor"))
measured.append(
(
character,
font_size,
font_name,
font_file,
color,
width,
background,
synthetic_italic,
)
)
total_width = sum(item[5] for item in measured)
align = str(props.get("align", "left"))
x = rect.x0
if align == "center":
x += max(0.0, (rect.width - total_width) / 2)
elif align == "right":
x += max(0.0, rect.width - total_width)
max_size = max((item[1] for item in measured), default=base_size)
line_box_height = max_size * line_height
background_top = y + max(0.0, (line_box_height - max_size) / 2)
background_height = max_size * 1.15
for (
character,
font_size,
font_name,
font_file,
color,
width,
background,
synthetic_italic,
) in measured:
if (
isinstance(background, str)
and background.lower() not in {"", "transparent", "none"}
):
page.draw_rect(
pymupdf.Rect(x, background_top, x + width, background_top + background_height),
color=None,
fill=_color(background),
fill_opacity=0.3,
overlay=True,
)
if character != " ":
options: dict[str, Any] = {
"fontsize": font_size,
"fontname": font_name,
"color": color,
"overlay": True,
}
if font_file:
options["fontfile"] = str(font_file)
origin = pymupdf.Point(x, y + font_size)
if synthetic_italic:
options["morph"] = _synthetic_italic_morph(origin)
page.insert_text(origin, character, **options)
x += width
y += max_size * line_height
return True
def render_text(page: pymupdf.Page, annotation: Annotation, _assets_dir: Path) -> None:
rect = _rect(annotation)
if rect is None:
return
props = _props(annotation)
text = str(props.get("text", ""))
if not text:
return
styles = props.get("styles")
highlight = props.get("highlightColor")
has_highlight = (
isinstance(highlight, str)
and highlight.strip().lower() not in {"", "transparent", "none"}
)
if has_highlight or (isinstance(styles, Mapping) and bool(styles)):
_render_styled_text(page, rect, text, props, styles)
return
align = {"left": 0, "center": 1, "right": 2}.get(str(props.get("align", "left")), 0)
font_size = max(1.0, float(props.get("fontSize", 14)))
font_name, font_file = _font_choice(
props.get("fontFamily"), props.get("bold"), props.get("italic")
)
synthetic_italic = _needs_synthetic_italic(
props.get("fontFamily"), props.get("bold"), props.get("italic")
)
options: dict[str, Any] = {
"fontsize": font_size,
"fontname": font_name,
"color": _color(props.get("color")),
"align": align,
"overlay": True,
}
if font_file:
options["fontfile"] = str(font_file)
if synthetic_italic:
options["morph"] = _synthetic_italic_morph(
pymupdf.Point(rect.x0, rect.y0 + font_size)
)
result = page.insert_textbox(rect, text, **options)
if result < 0:
_insert_text_lines(
page,
rect,
text,
fontsize=font_size,
font_name=font_name,
font_file=font_file,
color=_color(props.get("color")),
align=align,
line_height=max(0.5, float(props.get("lineHeight", 1.2))),
synthetic_italic=synthetic_italic,
)
def render_draw(page: pymupdf.Page, annotation: Annotation, _assets_dir: Path) -> None:
rect = _rect(annotation)
if rect is None:
return
props = _props(annotation)
svg_path = props.get("svgPath")
if isinstance(svg_path, str) and _draw_svg_path(
page,
rect,
svg_path,
color=_color(props.get("strokeColor")),
width=float(props.get("strokeWidth", 2)),
opacity=_opacity(props.get("opacity")),
):
return
_draw_polyline(
page,
_points_from_props(annotation, rect),
color=_color(props.get("strokeColor")),
width=float(props.get("strokeWidth", 2)),
opacity=_opacity(props.get("opacity")),
)
def _asset_path(assets_dir: Path, ref: Any) -> Path | None:
if not isinstance(ref, str):
return None
try:
safe_ref = str(uuid.UUID(ref))
except ValueError:
return None
path = assets_dir / safe_ref
return path if path.is_file() else None
def render_image(page: pymupdf.Page, annotation: Annotation, assets_dir: Path) -> None:
rect = _rect(annotation)
asset = _asset_path(assets_dir, _props(annotation).get("ref"))
if rect is None or asset is None:
return
page.insert_image(rect, filename=str(asset), keep_proportion=False, overlay=True)
def render_signature(page: pymupdf.Page, annotation: Annotation, assets_dir: Path) -> None:
rect = _rect(annotation)
props = _props(annotation)
if rect is None:
return
if props.get("mode") == "draw":
asset = _asset_path(assets_dir, props.get("ref"))
if asset is not None:
page.insert_image(rect, filename=str(asset), keep_proportion=False, overlay=True)
return
text = str(props.get("text", ""))
if text:
font_file = _signature_font_file(props.get("fontFamily"))
font_name = _font_name(props.get("fontFamily"), False, False)
font_size = _signature_font_size(
rect,
text,
font_file,
font_name,
props.get("fontSize"),
)
font = (
pymupdf.Font(fontfile=str(font_file))
if font_file
else pymupdf.Font(fontname=font_name)
)
options: dict[str, Any] = {
"fontsize": font_size,
"color": _color(props.get("color")),
"overlay": True,
}
# The font carries its own negative script-glyph bearings, so the PDF
# insertion origin is the Fabric text box edge. Fit the run to the
# canonical width to account for browser/PDF shaping differences and
# for intentional non-uniform resizing.
text_x = rect.x0
text_baseline = rect.y0 + font.ascender * font_size
text_width = float(font.text_length(text, fontsize=font_size))
if text_width > 0 and rect.width > 0:
options["morph"] = (
pymupdf.Point(text_x, text_baseline),
pymupdf.Matrix(rect.width / text_width, 0, 0, 1, 0, 0),
)
if font_file:
options["fontname"] = f"paperjet-{font_file.stem.lower()}"
options["fontfile"] = str(font_file)
else:
options["fontname"] = font_name
page.insert_text(pymupdf.Point(text_x, text_baseline), text, **options)
def render_highlight(page: pymupdf.Page, annotation: Annotation, _assets_dir: Path) -> None:
rect = _rect(annotation)
if rect is None:
return
props = _props(annotation)
page.draw_rect(
rect,
color=None,
fill=_color(props.get("color"), (1.0, 0.92, 0.1)),
fill_opacity=_opacity(props.get("opacity"), 0.3),
overlay=True,
)
def render_shape(page: pymupdf.Page, annotation: Annotation, _assets_dir: Path) -> None:
rect = _rect(annotation)
if rect is None:
return
props = _props(annotation)
kind = str(props.get("kind", "rect"))
stroke = _color(props.get("strokeColor"))
fill_value = props.get("fillColor")
fill = None if fill_value in (None, "", "transparent", "none") else _color(fill_value)
width = max(0.1, float(props.get("strokeWidth", 2)))
if kind == "ellipse":
page.draw_oval(rect, color=stroke, fill=fill, width=width, overlay=True)
elif kind in {"line", "arrow"}:
start = props.get("start")
end = props.get("end")
if (
isinstance(start, Sequence)
and not isinstance(start, (str, bytes))
and len(start) == 2
and isinstance(end, Sequence)
and not isinstance(end, (str, bytes))
and len(end) == 2
):
try:
start_point = pymupdf.Point(
rect.x0 + float(start[0]) * rect.width,
rect.y0 + float(start[1]) * rect.height,
)
end_point = pymupdf.Point(
rect.x0 + float(end[0]) * rect.width,
rect.y0 + float(end[1]) * rect.height,
)
except (TypeError, ValueError):
start_point, end_point = rect.tl, rect.br
else:
start_point, end_point = rect.tl, rect.br
page.draw_line(start_point, end_point, color=stroke, width=width, overlay=True)
if kind == "arrow":
angle = math.atan2(end_point.y - start_point.y, end_point.x - start_point.x)
length = min(
12.0,
max(
5.0, math.hypot(end_point.x - start_point.x, end_point.y - start_point.y) * 0.2
),
)
left = pymupdf.Point(
end_point.x - length * math.cos(angle - math.pi / 6),
end_point.y - length * math.sin(angle - math.pi / 6),
)
right = pymupdf.Point(
end_point.x - length * math.cos(angle + math.pi / 6),
end_point.y - length * math.sin(angle + math.pi / 6),
)
page.draw_polyline([left, end_point, right], color=stroke, width=width, overlay=True)
else:
page.draw_rect(rect, color=stroke, fill=fill, width=width, overlay=True)
HANDLERS: dict[str, Handler] = {
"text": render_text,
"draw": render_draw,
"signature": render_signature,
"image": render_image,
"highlight": render_highlight,
"shape": render_shape,
}
def export_annotations(
pdf_path: Path,
annotations: Sequence[Annotation],
assets_dir: Path,
) -> bytes:
"""Return a flattened PDF while preserving the source page rotations."""
with pymupdf.open(pdf_path) as document:
rotations = [page.rotation for page in document]
try:
for page in document:
if page.rotation:
page.set_rotation(0)
page_annotations: dict[int, list[Annotation]] = {}
for annotation in annotations:
try:
page_number = int(annotation.get("page", -1))
except (TypeError, ValueError):
logger.warning("Skipping annotation with invalid page: %r", annotation)
continue
if 0 <= page_number < len(document):
page_annotations.setdefault(page_number, []).append(annotation)
else:
logger.warning("Skipping annotation outside document pages: %r", annotation)
for page_number, page in enumerate(document):
ordered = sorted(
page_annotations.get(page_number, []),
key=lambda item: int(item.get("z", 0)),
)
for annotation in ordered:
annotation_type = annotation.get("type")
handler = HANDLERS.get(str(annotation_type))
if handler is None:
logger.warning("Skipping unsupported annotation type %r", annotation_type)
continue
try:
handler(page, annotation, assets_dir)
except Exception:
logger.exception("Skipping malformed %s annotation", annotation_type)
finally:
for page, rotation in zip(document, rotations, strict=True):
if page.rotation != rotation:
page.set_rotation(rotation)
return document.tobytes(garbage=4, deflate=True)